Plain-English transparency on how Earling Labs products use AI — which features run on Claude (made by Anthropic), which run on our own proprietary models, and what that means for you as a user.
Earling Labs builds AI-powered software for trade contractors, small-business operators, and active traders. Most of our AI-powered features — drafting estimates, summarizing leads, drafting payment reminders, extracting receipt data, drafting review requests — are powered by Claude, an AI assistant developed by Anthropic, a U.S.-based AI safety company. A smaller number of our features run on proprietary machine-learning models built by Earling Labs (most notably, the Earling Labs Predictor used for stock forecasting in our Trading Platform).
AI is a tool that drafts content and assists with workflow. Final decisions and final outputs are always yours. Every AI-generated draft is reviewable, editable, and never sent on your behalf without your explicit action.
AI may produce content that is inaccurate, incomplete, or misleading. Always review AI-generated content before sending it to customers, signing it on someone else's behalf, or relying on it for important decisions. You are responsible for the final content sent to your customers — Earling Labs products are tools that draft content; you decide what gets sent.
Here's a transparent breakdown of which features in each Earling Labs product use AI, and what kind of AI powers them:
| Product | AI Features | Powered By |
|---|---|---|
| Ringer | AI summary of homeowner lead messages for SMS alerts to contractors | Claude |
| Estimator | AI-drafted estimates by trade · photo measurement (Vision) · blueprint extraction (Vision) · suggested pricing from historical data | Claude |
| Scheduler | AI-drafted appointment confirmations + reminders | Claude |
| Invoicer | AI-drafted payment reminders in friendly / firm / final tones | Claude |
| Receipts | Receipt data extraction from photos (Vision) — vendor, total, date, category, payment method, with per-field confidence ratings | Claude |
| Booster | AI-drafted Google review-request emails in friendly / brief / formal tones | Claude |
| JOBS | Suggested inspection checklists from a job's trades · permit document reading (suggests permit fields for you to review) · read-only help bar that answers questions about your jobs. Nothing is saved until you confirm. | Claude |
| Reporter | Plain-English summaries of business figures that Reporter calculates itself (the AI describes the numbers; it does not produce them) · read-only help bar. Only aggregated figures are sent — no customer names or line items. | Claude |
| Crew | Read-only help bar that answers questions about your own roster, document expiry status, assignments, and hours. It cannot add, edit, or assign anything. | Claude |
| Routes | Read-only help bar for questions about your routes and stops · · · Stop ordering uses a road-distance routing engine (OpenRouteService) — not generative AI |
Claude Routing engine |
| Designer | Read-only help bar · · · Before/after preview renders come from a separate image-edit vendor, not Claude (until one is connected, a clearly-labeled simulated recolor is shown). Renders are illustrative previews, not a guarantee of the finished work. |
Claude Image vendor |
| Bookkeeper | Suggested tax categories for bank transactions (merchant, amount, and the bank's category hint) for you to review · help bar for questions about your own books. Not tax advice; never moves money. | Claude |
| Marketer | AI-drafted Google Business Profile posts · local-SEO keyword ideas · help bar. Nothing is published without your click; the listing score is a plain calculation, not AI. | Claude |
| Closer | AI-drafted reactivation texts and emails to your own past customers and open estimates · help bar. Nothing sends until you click Send; phone numbers and email addresses are not sent to the AI, and opt-out handling is fixed rules, not AI. | Claude |
| Trading Platform (in development) |
AI research chat (analysis, gut-checks, journaling) · · · Earling Labs Predictor — multi-horizon quantile forecasts (1d/3d/5d) for stock price movements |
Claude Proprietary ML |
Claude is an AI assistant developed by Anthropic, a U.S.-based AI safety company. When you use AI features in our products, your prompts (and the resulting outputs) are processed through Anthropic's API. Anthropic is a third-party service we license like any other developer would — we are not affiliated with Anthropic beyond being one of their many API customers.
The Earling Labs Predictor is a machine-learning model we built, trained, and host ourselves. It is not powered by Claude or any external AI service. Predictor uses quantile regression to forecast multi-horizon stock price ranges from historical price + volume data. The model weights, training pipeline, and inference code are owned by Earling Labs.
AI-generated content — every estimate, every payment reminder, every review-request email, every receipt extraction — is a draft. You see it, you can edit it, you decide whether and when to send. Nothing is sent on your behalf without your action. Take the moment to review. AI is a powerful drafting assistant, but it makes mistakes — wrong numbers, hallucinated facts, misremembered context, awkward phrasing for a particular customer.
When you click "Send" on an AI-drafted estimate, invoice reminder, or review request, the content goes out under your name to your customer. You — not Earling Labs, not Anthropic — are responsible for what was sent. Earling Labs provides the tool; you provide the judgment.
To provide AI features, the data you put into Earling Labs products (your prompts, your photos for Vision, your customer details for AI-drafted emails) may be sent to Anthropic's API for processing. We do not sell your data or your customers' data to third parties. Anthropic's own policies on data retention and training are available in their Privacy Policy. For details on how Earling Labs handles your data overall, see our Privacy Policy.
If you ever want to know how a specific feature works, whether it uses Claude or proprietary ML, or what happens to a specific piece of data, email support@earlinglabs.com. We respond personally. Transparency on AI is a value we take seriously — we'd rather over-disclose than have you surprised.
We are transparent about what we built and what we license. We do not claim to have built "our own AI" for features that are powered by Claude. We do not claim our products use "proprietary AI" when the language generation is coming from Anthropic. The only Earling Labs feature that is genuinely our own AI is the Earling Labs Predictor — and we'll always be clear about which features use Predictor vs. Claude.
If you ever see Earling Labs marketing or product copy that overstates what's proprietary vs. external — that's a mistake. Please point it out. We'll fix it.